Automate Everything
With n8n AI Agents.
No code, no experience needed - go from your first node to multi-agent workflows.
×
n8n Partner
Scan to get the slides.
Open them on your own device and follow every click - and revisit anytime during the week.
n8n-bootcamp.avanai.io
Four days, from your first node to a real AI-powered automation.
n8n Foundations
- Canvas, data & expressions
- Manual & Chat triggers
- Your first workflow
- Your first AI Agent (MGA)
- FLOW AI Agent demo
You leave having built & run your first workflow and chatted with an agent.
n8n Core
- Triggers: Schedule, Webhook, Form
- HTTP, credentials & Databricks
- IF, Switch, Filter, Merge
- Loop, Split Out, Aggregate
- The Code node
You can pull data from anywhere and shape it into exactly what you need.
n8n Expert
- Files: .csv, PDF, Sheets
- Errors, retries & microflows
- Datatables & caching
- AI agents: tools & memory
- Evals, guardrails & multi-agent
You can handle files, make flows robust, and give agents tools & memory.
Capstone: Invoice Automation
- Build the Invoice Reminder
- Find due invoices (Databricks)
- AI-drafted reminders
- Send via MS Outlook
- Open Q&A · ends 15:00
You ship a real, scheduled, AI-powered invoice reminder - end to end.
Each day: 3h morning · lunch 12:00-13:00 · 3h afternoon · a coffee break in each half · the idea fishbowl opens every afternoon.
n8n Foundations
Your gentle on-ramp: see what n8n can do, then build your first workflow and your first AI agent.
Welcome & what n8n is
The glue between your systems and AI: the canvas, nodes, connections, and the execution view.
Live demos
Watch finished n8n + AI automations run end to end - where we're headed this week.
Coffee break
Willi
Placeholder - to be defined.
Lunch
Idea fishbowl #1
Open round-table - spot the tasks worth automating in your own role.
n8n 101: canvas, data & expressions
Hands-on: items & JSON, the node detail view, expressions, and the Manual trigger - build your first workflow.
Coffee break
Your first AI Agent (MGA)
Add an AI Agent with the MGA chat model, give it one tool, and chat with it live.
FLOW AI Agent demo
A look at a richer AI Agent workflow - a taste of the agent skills coming on Day 3.
n8n Core
The core toolkit: get data in from anywhere with triggers and HTTP, then shape it with the essential nodes.
Recap & today's plan
A quick recap of Day 1 and what we'll cover today.
Triggers in depth
How workflows start: Schedule, Webhook, Form and app triggers, plus Respond to Webhook.
Coffee break
HTTP, credentials & Databricks
The HTTP Request node, API-key & OAuth credentials, and pulling real data from Databricks.
Lunch
Idea fishbowl #2
Pressure-test which of your tasks are a good fit for n8n.
Logic & data: IF, Switch, Filter, Merge
Branch, filter and combine data so each item flows the right way.
Coffee break
Looping, transforming & Code
Loop Over Items, Split Out, Aggregate, Sort/Limit/Remove Duplicates, Date & Time, and the Code node.
n8n Expert
Level up: files and formats, making flows robust, and giving AI agents tools, memory and structure.
Recap & plan
A look back at Day 2 and today's expert toolkit.
Files & formats
Extract from File (.csv, PDF, JSON), Read/Write Files, Convert to File, Spreadsheet and HTML/Markdown.
Coffee break
Robustness, microflows & Datatables
Error Trigger, Stop and Error, Continue On Fail & retries; reusable sub-workflows; Datatables & caching.
Lunch
Idea fishbowl #3
Where could agents, memory or stored data help in your work?
AI Agents: tools, memory & output
Give agents tools (HTTP/Code/Workflow), add Memory, and force clean results with a Structured Output Parser.
Coffee break
Evals, guardrails & multi-agent
Showcase: evaluate your AI outputs for quality, add guardrails and approvals, and have agents call other agents.
Capstone: Invoice Automation
Put it all together: build a real, scheduled, AI-powered invoice-reminder workflow, end to end.
Recap & the build plan
The capstone: a scheduled workflow that emails reminders for invoices about to be due.
Capstone Pt 1 · data in
Schedule Trigger, query invoices from Databricks (HTTP), and filter the ones due soon with Date & Time.
Coffee break
Capstone Pt 2 · AI & send
An AI Agent (MGA) drafts a personalised reminder per invoice, then send it via MS Outlook - made error-proof.
Lunch
Idea fishbowl #4
Turn the week's ideas into the automation you'll build next.
Coffee break
Open Q&A on your builds
Bring your own workflow ideas and blockers - we'll work through them together.
Wrap-up & close
Key takeaways, resources, and where to go next.
n8n 101: canvas, data & expressions
Your first hands-on build - get comfortable with the canvas, see how data flows from node to node, write your first expressions, then build and run a workflow end to end.
🧰 You'll use: items & JSON · the node detail view · expressions · the Manual trigger · the Set node
n8n 101: canvas, data & expressions
New hands-on session, new AvaStage code - hop in to follow every click live on your own screen.
Your first AI Agent (MGA)
Add an AI Agent powered by Bayer's MGA chat model, give it a single tool to call, and chat with it live - your first taste of AI working inside an n8n workflow.
🧰 You'll use: the AI Agent node · the MGA chat model · one tool · the Chat trigger
Your first AI Agent (MGA)
New hands-on session, new AvaStage code - hop in to follow every click live on your own screen.
Triggers in depth
Every workflow needs a way to start. Explore the main triggers - on a schedule, from a webhook, from a form, or from an app event - and reply to callers with Respond to Webhook.
🧰 You'll use: Schedule · Webhook · Form · app triggers · Respond to Webhook
Triggers in depth
New hands-on session, new AvaStage code - hop in to follow every click live on your own screen.
HTTP, credentials & Databricks
Reach any system on the internet - call an API with the HTTP Request node, keep secrets safe with credentials, and pull real data live from Databricks.
🧰 You'll use: the HTTP Request node · API-key & OAuth credentials · Databricks
HTTP, credentials & Databricks
New hands-on session, new AvaStage code - hop in to follow every click live on your own screen.
Logic & data: IF, Switch, Filter, Merge
Make decisions inside your flow - branch on a condition, route many ways, drop items you don't want, and recombine streams so each item takes exactly the right path.
🧰 You'll use: IF · Switch · Filter · Merge
Logic & data: IF, Switch, Filter, Merge
New hands-on session, new AvaStage code - hop in to follow every click live on your own screen.
Looping, transforming & the Code node
Reshape data at scale - loop over items, split and aggregate, sort, limit and dedupe, work with dates and times, and drop into the Code node when you need full control.
🧰 You'll use: Loop Over Items · Split Out · Aggregate · Sort/Limit/Remove Duplicates · Date & Time · Code
Looping, transforming & the Code node
New hands-on session, new AvaStage code - hop in to follow every click live on your own screen.
Files & formats
Get data into and out of files - read CSVs, PDFs and JSON, write files back out, and convert freely between spreadsheet, HTML and Markdown.
🧰 You'll use: Extract from File · Read/Write Files · Convert to File · Spreadsheet · HTML/Markdown
Files & formats
New hands-on session, new AvaStage code - hop in to follow every click live on your own screen.
Robustness, microflows & Datatables
Make flows production-ready - catch and handle errors, retry safely, factor logic into reusable sub-workflows, and store or cache state with Datatables.
🧰 You'll use: Error Trigger · Stop and Error · Continue On Fail & retries · sub-workflows · Datatables
Robustness, microflows & Datatables
New hands-on session, new AvaStage code - hop in to follow every click live on your own screen.
AI Agents: tools, memory & output
Level up your agents - give them tools to call (HTTP, Code, Workflow), add Memory so they keep context across turns, and force clean, predictable results with a Structured Output Parser.
🧰 You'll use: agent tools (HTTP/Code/Workflow) · Memory · Structured Output Parser
AI Agents: tools, memory & output
New hands-on session, new AvaStage code - hop in to follow every click live on your own screen.
Evals, guardrails & multi-agent
A showcase of where agents go next - judge your AI outputs for quality, add guardrails and human approvals, and let agents call other agents. Follow along live; we'll apply the basics in the capstone.
🔭 You'll see: AI evaluations · guardrails & approvals · multi-agent
Evals, guardrails & multi-agent
New hands-on session, new AvaStage code - hop in to follow every click live on your own screen.
Capstone Pt 1 · data in
Start the capstone, the Invoice Reminder - a Schedule Trigger kicks it off on its own, an HTTP call pulls invoices from Databricks, and Date & Time filters down to the ones due soon.
🧰 You'll use: Schedule Trigger · HTTP Request / Databricks · Date & Time filter
Capstone Pt 1 · data in
New hands-on session, new AvaStage code - hop in to follow every click live on your own screen.
Capstone Pt 2 · AI & send
Finish the capstone - an AI Agent (MGA) drafts a personalised reminder for each invoice, you send it via MS Outlook, and you make the whole workflow error-proof so it runs unattended.
🧰 You'll use: AI Agent (MGA) · MS Outlook (Graph) · error handling & a microflow
Capstone Pt 2 · AI & send
New hands-on session, new AvaStage code - hop in to follow every click live on your own screen.
Aemal Sayer
CTO & Co-Founder · Avanai
Avanai is an n8n expert partner, working with Bayer to deliver hands-on n8n enablement. I help teams go from zero to building real automations - exactly like the two you'll build today.
n8n is the glue between your systems and AI.
Think of every tool you use - email, databases, spreadsheets, approval systems, AI models - as an island. n8n is the bridge between them. You connect steps (called nodes) into a workflow, and n8n runs them in order, passing data from one to the next. That's the whole concept.
Nodes
Each box is one step - "fetch this data", "send that email", "ask the AI". You add them from the node panel and wire them together on the canvas.
Workflows
A chain of nodes from a trigger to an outcome. Data flows left to right. No buttons to press once it's running - it just works.
Executions
Every run is recorded. The execution view shows exactly what each step did and what data came back - your window into the workflow.
🧭 What n8n is not, today: it's not a tool for building forms or user interfaces. We use it to connect systems and AI - that's where its power is.
Every workflow begins with a trigger.
A trigger is the event that kicks a workflow off and hands it the first piece of data. Pick the one that fits the job - the rest of the nodes just follow. You'll use two of these today.
Chat / Manual
You start it - a chat message or a one-click run.
Schedule
Runs on a clock - every hour, every morning, every Monday.
App event
"New email", "row added", "form submitted" - n8n reacts on its own.
Webhook
Another system calls n8n directly with a URL to start the flow.
🔌 And it connects to everything: 400+ ready-made app integrations, 30+ AI model nodes (OpenAI, Claude, Gemini), and an HTTP node that can reach any API on the planet - so nothing is off-limits.
From a single email to a full AI agent.
The same building blocks cover a huge range of work. A few patterns teams automate every day:
Reports & alerts
Pull data on a schedule, format it, and email or Slack it - exactly like Workflow 2 today.
AI agents & chatbots
Answer questions over your own data, call tools, and take actions - like Workflow 1.
Support automation
Triage tickets, draft replies, look up orders, and escalate the tricky ones.
Data sync
Keep CRM, spreadsheets, and databases in step without copy-paste.
IT & security ops
Enrich alerts, automate routine fixes, and cut response time.
Document extraction
Read invoices, forms, and PDFs with AI and push the data where it belongs.
📈 Real results: Field Aerospace cut proposal drafting from two weeks to ~25 minutes. Koralplay now auto-resolves 70% of payment tickets. Vodafone saved £2.2M automating threat intelligence.
Built to run in a regulated enterprise.
The same tool you're learning today scales straight to production - which is why 34% of the Fortune 500 already run on n8n.
Security & control
Self-host or cloud, SSO / SAML, SOC 2, and secrets in Vault, AWS, or Azure - your data stays yours.
Governance
Project-level access, Git as the source of truth, and separate dev & production environments.
Confidence at scale
AI evaluations before you ship, 200+ executions/sec, and an insights dashboard to prove ROI.
🏢 In good company: Meta, Microsoft, Vodafone, and Zendesk build on n8n. Source: n8n.io/enterprise
From one workflow to millions of runs.
The workflow you build today on a laptop is the same workflow that runs in production - you don't rebuild it to scale, you just add capacity behind it.
Queue mode & workers
Hand executions to a pool of worker processes. Need more throughput? Add more workers - no workflow changes.
Multi-main & failover
Run several main instances with automatic failover, so a single node going down doesn't stop the line.
High throughput
200+ executions per second on a tuned setup - enough to drive real, business-critical volume.
🩺 And you can see it all: execution logs stream into your own monitoring stack, and the insights dashboard tracks runs, failure rates, and time saved as you scale up.
Every concept & node, and where you'll learn it.
Your week at node level - we'll say a few words on each and tick it off as we cover it.
| Concept / node | What it is | Day | |
|---|---|---|---|
| Canvas, NDV & executions | How n8n is structured, and how to read what a run did | Day 1 | |
| Data model & expressions | Items, JSON, and {{ }} to reference earlier data | Day 1 | |
| Triggers: Manual & Chat | Run on demand, or talk to it via a chat box | Day 1 | |
| AI Agent + Chat Model (MGA) | An agent that reasons, calls tools, and answers | Day 1 | |
| Triggers: Schedule, Webhook, Form | Start on a clock, an incoming event, or a form | Day 2 | |
| HTTP Request + credentials | Call any API; API-key & OAuth authentication | Day 2 | |
| Databricks (via HTTP) | Pull real invoice data from the warehouse | Day 2 | |
| Edit Fields, IF, Switch, Filter | Set values, then branch and filter items | Day 2 | |
| Merge | Combine data from several branches | Day 2 | |
| Loop, Split Out, Aggregate | Process a list item by item, then recombine | Day 2 | |
| Sort, Limit, Dedupe, Date & Time | The everyday data utilities | Day 2 | |
| Code node | Custom JavaScript when the built-in nodes aren't enough | Day 2 | |
| Files: Extract / Read / Write / Convert | Work with .csv, PDF, JSON, spreadsheets & HTML | Day 3 | |
| Error handling & retries | Error Trigger, Stop and Error, Continue On Fail | Day 3 | |
| Microflows & Datatables | Reusable sub-workflows; stored state & caching | Day 3 | |
| AI agents: tools, memory, output | Tools, Memory, and the Structured Output Parser | Day 3 | |
| Evals, guardrails & multi-agent | The advanced AI layer (showcase) | Day 3 | |
| MS Outlook send + the capstone | Email the reminders, and assemble the full build | Day 4 |
Let's see it run, end to end.
Before we build anything, here's the destination. A quick live run of both finished workflows - the AI Agent you can chat with, and the scheduled invoice email landing in the inbox.
Follow every click, live, on your own screen.
We'll use AvaStage to keep you in sync. It mirrors my screen to your device with a fresh screenshot every 3 seconds - miss a click? Rewind and replay it at your own pace. Got a question? Post it in AvaStage and upvote others'. I'll jump back in regularly to answer them, so don't hold your questions to the end - fire them in as they come.
Chat with an AI Agent
Your first workflow. An AI Agent that answers questions by calling a tool to fetch live data - and you'll talk to it.
⬇ Download workflow JSON
Chat in → agent thinks → agent calls a tool → agent answers.
message received"
The agent has one tool: get_users, which fetches a list of users from a free, public endpoint.
When you ask about users, the agent decides on its own to call the tool, reads what comes back, and answers you.
It's the simplest possible way to see tool calling in action.
Three nodes, wired in a couple of clicks.
get_usersGETjsonplaceholder.typicode.com/users🧪 Why this endpoint: it returns dummy users, needs no API key, and never fails in front of a live audience. It's just a clean vehicle to demonstrate tool calling.
Talk to your agent. Watch it reach for the tool.
🔎 The teaching moment: after each answer, open the execution view. You'll see the tool being called and the data coming back - proof the agent isn't guessing, it's fetching. This is the heart of the first 30 minutes.
Now let's build it - step by step.
Time to switch to the n8n canvas and build Workflow 1 together, node by node. Follow along in AvaStage - rewind any step you miss, and keep firing your questions as we go.
Scan the slidesBack at -
Up next · Workflow 2 · Fetch & clean the invoice data
Scheduled invoice email
The complete automation loop - data in, logic applied, an email sent - running on its own every morning. Nobody presses a button.
⬇ Download workflow JSON
Fetch → clean → send. Every morning at 9 AM.
This is the same fetch → clean → send pattern behind countless real automations - just with real-feeling invoice data in the middle. Once the schedule is on, the workflow runs itself: it pulls the invoices due soon, turns them into an email, and sends it. The "we could automate our invoice chasing" moment.
Four steps from a schedule to an inbox.
09:00📨 Each person sends to their own address - not a shared distribution list - so the email lands in a place you control and can open right away.
Every morning, your invoices are already triaged.
Once this workflow is live it runs by itself - every morning at 9 AM, before you even open your day. You just find the email waiting:
- 📊Top 10 invoices due in the near future, ranked for you.
- 🤖A short AI-generated summary of each one.
- 🗄️Pulled live from Databricks - reflecting your real invoice sources (SmartPay / SAP).
- ⏰No clicks, no chasing - it just lands in your inbox.
Now let's build it - step by step.
Time to switch to the n8n canvas and build Workflow 2 together, node by node. Follow along in AvaStage - rewind any step you miss, and keep firing your questions as we go.
Scan the slidesBack at -
Up next · Workflow 2 · Compose, send & run it live
Run it once. Watch the email arrive. Then let it loop.
▶️ What we do live
Now imagine your real data
Swap the demo source for your actual invoice, PO, or approval data and this same workflow becomes a vendor-reminder service, an approval nudge, or an overdue-invoice chaser - built by you, running on its own.
You just built the seed of a bootcamp workflow.
What you built today is a deliberately simplified version of a workflow from the week-long bootcamp. Same backbone - a schedule pulling invoice data and emailing it - but in the bootcamp it grows up.
✅ Today · the starter version
Schedule → Databricks → compose → send. The complete loop, end to end, built by a complete beginner.
🚀 In the bootcamp · the full version
The same workflow plus header enrichment, data tables, caching, and AI analysis of the invoices. Recognise the through-line.
Scan the slidesBack at -
Up next · Quiz & your ideas
Quick quiz
Five questions, multiple choice, just for fun. One of you answers - we reveal together. Confetti if you nailed it.
What is n8n best described as in today's session?
The answer
n8n is the glue between your systems and AI. It connects the tools you already use and passes data between them - that's the whole idea we built on today. It is not a UI or form builder.
In Workflow 1, what did the AI Agent use to fetch the list of users?
The answer
The agent called a tool - the get_users HTTP Request - to fetch the data, then answered
from what came back. You saw exactly that happen in the execution view.
What made Workflow 2 run by itself every morning?
The answer
The Schedule Trigger is what makes a workflow run on its own. Set it to daily at 09:00 and the whole chain fires every morning - no button, no person.
Where do you look to see exactly what each step of a workflow did?
The answer
The execution tab records every run and shows what each step did and what data came back. It's your window into the workflow - and where you saw the tool being called.
Which of these did we deliberately NOT use today?
The answer
We deliberately avoided Microsoft Teams - to keep from notifying real people we don't control. Each email went to your own inbox instead, so the lab stays safe and self-contained.
What could you automate?
You've now built both patterns - an AI Agent that calls a tool, and a scheduled job that fetches, shapes, and sends. Almost every repetitive task at Bayer PH is some mix of these two. A few starting points:
Invoice chasing
Email vendors automatically when an invoice is overdue.
Approval nudges
Remind approvers about purchase requests waiting on them.
Daily PO digest
A morning summary of new purchase orders, in your inbox.
Ask-your-data agent
An AI Agent that answers "which invoices are due this week?"
Exception alerts
Flag invoices over a threshold or missing a PO reference.
Vendor onboarding
Route a new-vendor form through the right steps automatically.
Q&A · open floor
Anything from today - the AI Agent, the scheduled email, n8n in general, or how this could fit your own work. Ask away.
You built two workflows.
Now go automate your job.
You'll get annotated, step-by-step guides for both - and there's a week-long bootcamp when you're ready to go deeper.
×
